In 2012, we reported a single case of a child with a pineoblastoma and a loss-of-function (LOF) germline DICER1 pathogenic variant [9]. Unlike all previously studied DICER1-related tumours, the second hit in the pineoblastoma was complete loss of the wild-type chromosome 14q (chr14q). Blood-derived cDNA analysis confirmed the absence of RNA product from the mutant allele. We then showed that this was not a unique case—in follow-up studies, all five pineoblastomas with DICER1 mutations for which tumour tissue was available also exhibited LOF DICER1 pathogenic variants paired with either another LOF variant or loss of heterozygosity (LOH) [1, 4], and in each case, we were able to show a complete absence of DICER1 protein by immunohistochemistry [1].

Three recent papers in Acta Neuropathologica confirm our earlier observations [4, 5, 7]. Combined, of the 41 pineoblastomas found to have alterations in or involving DICER1, 22 exhibited biallelic DICER1 inactivation.

In DICER1 syndrome, the predisposing alteration is most often LOF in nature. The second somatic events in DICER1 lesions are predominantly missense alterations involving one of five amino acid residues in the RNase IIIb domain (Fig. 1) [2]. These so-called “hotspot” mutations have been demonstrated to interfere with catalytic cleavage of one strand of a microRNA hairpin precursor, resulting in altered microRNA expression patterns that are hypothesized to underlie tumour formation. To answer the question of whether the mechanism of DICER1 alteration in pineoblastoma is truly distinct from other DICER1 syndrome-related tumours, we used a catalogue of all DICER1 alterations reported in the literature as of December 2019 (updated from [2]) to compare the prevalence of LOH and biallelic inactivation in DICER1 syndrome tumours.

Fig. 1
figure 1

Graphic representation of the different combinations of genetic events involving DICER1 observed in tumours. In the table, the total number of pineoblastomas bearing the depicted biallelic genetic events is indicated, along with that for all other tumour types combined. Superscript a: please refer to Online Resource 1 for a breakdown of the other genetic alterations that are not depicted here. Asterisk: P value for the difference between these two percentages is 8.18E−37 (Online Resource 2C)

Full size image

Of 668 tumours found to bear one or more DICER1 alterations (whether or not DICER1 syndrome related), 59 (9%) had LOH of the wild-type allele, focal deletions of DICER1, or homozygous loss of chr14q [i.e. complete loss of at least one DICER1 allele (hereafter, allele loss)] [Online Resource (OR) 1 and Fig. 1]. The tumours bearing allele loss were comprised of nine different types and prevalence of allele loss in each tumour type ranged from 2 to 67%. Allele loss was most prevalent in pineoblastomas (33/49, 67%), followed by DICER1-related central nervous system (CNS) sarcomas (8/35, 23%) (OR1); the prevalence of allele loss in each of these tumour types was statistically significantly greater compared to that in other tumour types combined, Fisher’s exact test p values < 0.0001 (OR2A–B).

Among the 350 tumours known to have two DICER1 alterations, 22 (6%) had an RNase IIIb hotspot mutation coupled with LOH; whereas 269 (77%) bore a hotspot mutation coupled with a LOF alteration—the latter being the classic pattern of mutations observed in DICER1 syndrome (Fig. 1 and OR3). In total, 32/350 (9%) tumours had two hits that truncated or removed the protein, 24 of which exhibited LOH coupled with a LOF DICER1 alteration (Fig. 1, OR1 and OR3). DICER1 function was thus majorly compromised or completely abrogated in these 32 tumours, which included 30 pineoblastomas, 1 Wilms tumour [6], and 1 breast carcinoma [12] (see OR2C–D for statistics). There are a further 318 tumours in which only 1 DICER1 alteration was identified, 256 of which were RNase IIIb hotpot mutation positive [2]. It remains possible that additional allele loss events have gone unrecognized in this set of tumours [10].

The evidence presented herein demonstrates that DICER1-related pineoblastomas predominantly exhibit complete DICER1 loss (30/49 cases, 61%; OR1), often bearing one LOF alteration (dispersed throughout the gene; OR4) coupled with LOH (23/30 cases; Fig. 1) [1, 3,4,5, 7, 9, 11]. Systematic analysis of the allele fractions contributing to the observed LOH was not performed in most of the published studies. Such analysis was performed by Pugh and colleagues on a set of 15 pleuropulmonary blastomas bearing DICER1 alterations [8]: 9/15 were copy quiet and 6/15 exhibited copy number gains or losses or copy-neutral LOH, notably, with retention or duplication of the RNase IIIb hotspot mutant allele. These findings suggest that allelic imbalance events in pleuropulmonary blastoma favour the retention of the allele bearing the RNase IIIb hotspot mutation. Further analyses are needed to determine whether this is also the case for other DICER1 syndrome lesions.

Biallelic inactivation of two other microRNA processing genes, DROSHA and DGCR8, have been identified in 31 and 6 pineoblastomas, respectively, which too perturb microRNA production [3, 5, 7, 11] (OR5; the prevalence of biallelic inactivation of DROSHA in pineoblastoma is statistically significantly greater than in Wilms tumour, a tumour type that frequently features microRNA biogenesis gene mutations, Fisher’s exact test p value < 0.00001; The numerical difference between biallelic inactivation of DGCR8 in pineoblastoma versus Wilms tumour is of borderline significance (p = 0.0103)). The pineoblastoma cell-of-origin thus appears uniquely capable of tolerating complete or near-complete absence of microRNAs generated through the canonical pathway.

DICER1-related CNS sarcomas also exhibit increased prevalence of LOH of DICER1 (typically coupled with an RNase IIIb hotspot mutation)—of note is that only 5/13 reported cases with available matched-normal DNA occurred in the presence of a germline DICER1 alteration, meaning that most cases are wholly post-zygotic in origin. LOH is significantly more prevalent in tumours bearing biallelic confirmed somatic DICER1 alterations than in tumours from patients with germline predisposing alterations (p = 1.6E−06; OR6), possibly explaining the increased prevalence of LOH in DICER1-related CNS sarcomas, and, in contrast to pineoblastoma, suggesting an intolerance of their normal cell-of-origin to near-complete loss of microRNA biogenesis. Thus, the order of DICER1 hits, their provenance, and the tissue of origin all appear to contribute to the likelihood of LOH being the second event in DICER1 syndrome tumours.

These findings illustrate that although allele loss and biallelic inactivation of DICER1 are infrequent mechanisms of alteration in DICER1 syndrome-related tumours, they do occur with increased prevalence in certain tumour types, particularly pineoblastsoma. An appreciable fraction of pineoblastomas are driven by microRNA processing alterations that cause global microRNA dysregulation. The apparent tolerance of many pineoblastomas for biallelic inactivation of one of the microRNA biogenesis genes presents a unique therapeutic opportunity that could be investigated with the use of a synthetic lethal screen of isogenic pineoblastoma cell lines with and without engineered microRNA biogenesis gene knockouts.